Big banks are blocking AI agents from logging into your accounts

Four of the biggest US banks are shutting the door on outside AI agents. Citi, Wells Fargo, Bank of America and JPMorgan Chase have all moved to limit them, and others are following. An AI agent is software that acts for you, such as logging in, checking balances or paying for something. Wells Fargo's new terms say you are responsible for what it does.

What changed

Citi says it does not let third-party AI agents log into a customer's account or make transactions on its digital banking platforms. Customers can still allow limited, read-only access through certain Citi-approved open banking providers. Bank of America is restricting agent access to consumer deposits. JPMorgan Chase's CEO told Bloomberg it is blocking outside agents from seeing customer-account information. Wells Fargo says it has taken steps to block unauthorized agents.

The Wells Fargo fine print

Wells Fargo has updated its online access agreement, effective Nov. 24. It says users are "solely responsible" for supervising any AI agent they use. Transactions made by an agent count as authorized, including ones you "did not want or intend." The exception is if you tell the bank a transaction was not authorized and it has had reasonable opportunity to act. Users also accept the security risk of letting an agent in.

Why banks are worried

Bank of America CEO Brian Moynihan put it simply: when you give an agent your credentials, you are the one authorizing the payment. His ask is a human in the loop, so you approve before you pay. Bank of America, Capital One and four other global banks formed a coalition in September warning that agent technology is moving faster than the rules meant to protect people. They flagged AI shopping agents in particular.

A 2026 Palo Alto Networks study found two-thirds of security chiefs in financial services and software ranked agentic AI among their top three cyber risks. Charles Schwab's CEO said the firm is blocking "tens and twenties" of attempts by major AI agents to reach client funds.

What is still unclear

Citi says it has not seen reports of agents accessing client accounts. Ping Identity CEO Andre Durand said he knew of no specific cases of agents hacking banks, but added he does not think much is being reported. His bigger question is how a bank can tell an agent from a human, and whether the agent truly has the person's permission. Bank of America and JPMorgan Chase did not immediately respond to USA Today.

What to do

Consumers seem open to AI help with money. A Ping survey found only 22% refuse to trust AI with any financial transaction, though 71% worry an agent could misrepresent their values. Durand says one-time codes sent by email are no longer safe enough, because agents often have access to email. He recommends:

• Use passkeys and biometrics, which confirm a human is signing in.
• Keep a human approval step before any payment.
• Read your bank's terms before sharing a login with any AI tool.